[apeg]

I have been doing loads of reading and cant seem to find a solid answer on the cachemembuffersize value on streaming video content.
(I have to Pi 512) I was having great success when I setting the value to 120971520 (128mb) on 720p videos (steaming) but on larger 10gb+ 1080p movies it does not seem to be cutting it.

I have no problem waiting at the very start of playing a video, ideally I would like it to buffer as much as needed so the movie will not get interrupted (yes I know on a slow connection that could mean downloading almost the entire movie) but im fine with that

Seeing how the Pi only has 512mb of ram (I’ve read that the cachemembuffersize is actually tripled) and that explains why I can increase to 241943040 (256mb)… but this isn’t important as 256, 512 all falls short of what I need

So, this brought me to <cachemembuffersize>0<…

From what I have been reading this pulls the buffer away from ram and uses the SD card, FANTASTIC! This is exactly what I want.
But I noticed on my large videos it tends to hang @ around 26% and never moves. My guess is the SD buffer has filled up and it doesn’t know what to do.

FINALLY my question:

If I have a 16gb SD card is it possible to increase the partition size on where ever the buffer is kept in order to handle my larger 10GB+ movies?

Cheers and thanks,
-aPeg

[Milhouse]

cachemembuffersize is likely to become deprecated if this patch ever lands so I wouldn't waste too much time tinkering with it. I believe the updated network buffering patch is already in the Rbej OpenELEC custom builds (see OpenELEC thread) so you could try that build and see if there is any improvement. I've no idea if you're using OpenELEC or not as you never bothered to mention your Pi distribution which is often one of the most vital pieces of information you can impart when reporting a problem.

[popcornmix]

Where are you streaming from?
cachemembuffersize is ignored for local content (i.e. SMB/NFS/USB).
It only applies to internet urls (e.g. starting with http://)

[spjonez]

cachemembuffersize is likely to become deprecated if this patch ever lands so I wouldn't waste too much time tinkering with it. I believe the updated network buffering patch is already in the Rbej OpenELEC custom builds (see OpenELEC thread) so you could try that build and see if there is any improvement. I've no idea if you're using OpenELEC or not as you never bothered to mention your Pi distribution which is often one of the most vital pieces of information you can impart when reporting a problem.

I've streamed a 17gb 3d 1080p DTS movie from one Pi (512) running Raspbian to another Pi (512) running rbej's OpenELEC build over ethernet (samba share) without issues. The cache system in his build works wonders for large files.

[apeg]

You're right, idk how i left that out:
i am using OpenELEC, image r13541 to be specific. I am streaming over the internet via SFTP, im sure the cachemembuffersize is used in my case because i see big differences as i play with the number.

Ill give Rbej OpenELEC custom build a shot and go from there.

Any idea how the new patch works? There is no fake way around the problem im hitting, does the new patch use the entire space of the SD card when needed?

-aPeg

[apeg]

dont get me wrong, im sure this new buffer patch is better for most situations, but it seems like a step back for my needs.

alwaysforcebuffer is nice for people running over a local network but in my case the buffer was already being used.

freememorycachepercent i guess is nice seeing how you don't have to calculate bit numbers yourself, but that still wont get me anywhere near the buffer size i need. And is there no way to offset the buffer to the SD card anymore?

[apeg]

got it,

new patch aside, i set the cachemembuffersize to 0 and re-sized the /dev/storage partition on the SD card to use all available space and we're up and running perfectly.

on a 13gb movie it takes a good 10min buffering at the start but than the entire movie plays smooth as butter Start > End (buffer got as large as 1.7gb before playing)

Thanks,
-aPeg

if the new patch makes it into the trunk i hope there is a way to mimic the buffer push to the SD card

[popcornmix]

if the new patch makes it into the trunk i hope there is a way to mimic the buffer push to the SD card

You should voice your concerns here:
https://github.com/xbmc/xbmc/pull/1388